How Language Shapes the Way You Think

Does the language you speak change how you think? For much of the 20th century, the dominant answer in linguistics was no: thought comes first, and words are just the delivery truck. Over the past few decades, experiments on color, direction, time, numbers, and decisions have pushed that answer toward a qualified yes. Language does not imprison thought, but it does seem to bias what people notice, remember, and find easy.

The qualifier matters as much as the yes. Some of the most famous results in this field have been challenged, and the honest summary is "real but smaller than the headlines suggest." This article walks through the best-documented findings, the failed replications, and what the evidence can and cannot tell you.

Key takeaways

• The strong Sapir-Whorf hypothesis (language determines thought) is rejected by most researchers. The weak version (language influences thought) has experimental support.
• Color: Having separate words for two shades can speed discrimination between them, especially in the right visual field. Some follow-up studies disagree.
• Space: Speakers of languages that use compass directions instead of left and right keep track of cardinal directions far more closely, and spatial habits shape how they arrange time.
• Numbers: The Pirahã language has no exact number words, and speakers struggle with exact quantities when memory is involved, which suggests number words work as a mental tool.
• Decisions: Wording changes choices in any language, and using a foreign language can reduce some of those effects.
• Several celebrated findings, including the Mandarin vertical-time result, have failed or mixed replications.

The Sapir-Whorf hypothesis: history and controversy

The idea that language shapes thought is linked to the linguist Edward Sapir (1884 to 1939) and his student Benjamin Lee Whorf (1897 to 1941). Neither man ever stated it as a formal hypothesis. The label "Sapir-Whorf hypothesis" was coined later by Harry Hoijer, and in 1954 Roger Brown and Eric Lenneberg formulated testable versions, according to the reference summary of linguistic relativity.

Whorf's best-known claim was that the Hopi language lacks the kind of time units English has, and so Hopi speakers experience time differently. In 1983, Ekkehart Malotki reported finding no support for that claim in 1980s Hopi speech or in historical documents. Some relativists, including John Lucy, argued Malotki misread Whorf, so the Hopi case remains disputed, but few researchers now treat it as the foundation of the field.

Two versions with very different odds

The strong form, linguistic determinism, says language fully determines what you can think. It was never really advanced by Sapir or Whorf, was attributed to them afterward, and has fallen out of favor. If it were true, you could not think a thought you lacked words for, and translation or learning new concepts would be impossible.

The weak form, usually called linguistic relativity, claims that language structures influence perception, memory, and reasoning without strictly limiting them. That is a much easier claim to defend, and it is the one researchers now test. The questions are narrow: how large is the effect, how robust is it, and does language cause it or merely correlate with it?

Think of language as a set of well-worn paths across a field. You can walk anywhere, but you tend to follow the paths. The research asks how strong that tendency is, and whether it comes from the paths themselves or from the kind of people who built the paths.

Color: where words may sharpen what you see

The color spectrum is continuous, but languages cut it into different numbers of named categories. Brent Berlin and Paul Kay's 1969 survey found that these categories follow universal patterns: languages with three basic color terms always have black, white, and red, for example. That finding cuts against extreme relativism, because it suggests perception constrains naming.

Early tests of the opposite direction looked at whether naming constrains perception. Lenneberg and Roberts found that Zuni speakers, who group orange and yellow into one category, had more trouble recalling those colors than English speakers did, and Brown and Lenneberg tied this to how easily colors are named.

The Russian blues experiment

Russian has two basic words for blue: goluboy (light blue) and siniy (dark blue). English has one. In a 2007 study published in PNAS, Jonathan Winawer, Lera Boroditsky, and colleagues asked Russian and English speakers to pick which of two color squares matched a target. Russian speakers answered faster when the two squares fell into different Russian categories than when both fell in the same one. English speakers showed no such pattern with the same stimuli. The paper is available on the Stanford Language and Cognition Lab site.

Two further details make the result more interesting. A verbal interference task, in which participants silently rehearsed a string of digits, removed the Russian advantage, while a spatial interference task did not. That suggests the effect depends on language being active in the moment. The effect was also strongest for shades that were hard to tell apart.

Right visual field, left hemisphere

The hemisphere result is subtle. Each half of the visual field is processed first by the opposite side of the brain, and language is usually handled in the left hemisphere, which receives input from the right visual field. In 2006, Aubrey Gilbert, Terry Regier, Paul Kay, and Richard Ivry reported that color-category effects appeared in the right visual field but not the left. A follow-up by Drivonikou and colleagues in 2007 confirmed the stronger right-field effect, but also found significant, weaker effects in the left field, according to the paper's abstract on PubMed Central.

The authors of that follow-up offered two readings. Linguistic information might cross between hemispheres and arrive weakened, or the left-field effect might reflect universal color categories already present in infants. Either way, the original claim that the effect vanishes entirely on the left was too strong.

A caution on replication

The Russian blues result has not gone unchallenged. A later study by Martinovic, Paramei, and MacInnes reported that the discrimination advantage at the Russian blue boundary was not reflected in processing speed, which contradicts the 2007 result. Commentators also pointed out that the findings are consistent with language shaping perception but do not by themselves prove it.

For the reader, the right takeaway is that color perception is probably not "fixed" by language, but categorical labels may add a small, fast, and context-dependent boost. Our article on the psychology of color covers other claims that sound more certain than the evidence allows.

Space: north, south, east, and west instead of left and right

English speakers describe location relative to their own bodies: the cup is to my left. Other languages use absolute directions. Guugu Yimithirr, spoken in Queensland, Australia, is notable for using cardinal directions instead of left and right, and its speakers are described as having a highly developed sense of direction. The claim of "pure" cardinal use has been disputed, as the reference page for the language notes, but the broad pattern is well documented.

If your language forces you to say "the cup is to the north of the plate," you have to know where north is at all times. That is a cognitive habit forced by grammar. Majid, Levinson, and colleagues argued in a 2004 review that spatial frames of reference, once assumed to be built into human neurocognition, vary across cultures, which is evidence that language can restructure spatial cognition (Majid et al., 2004).

Time laid out along the compass

Lera Boroditsky and Alice Gaby tested how speakers of Kuuk Thaayorre, from the community of Pormpuraaw in Australia, arrange pictures of temporal progression, such as a person aging. They reported in 2010 that time ran along an east-to-west axis for these speakers, regardless of which way they faced. Follow-up work by Gaby reported a similar east-to-west pattern, which the literature contrasts with the left-to-right arrangement typical of English speakers.

Here is how to read that. English speakers likely arrange time left to right partly because of reading and writing direction, and partly because of how English describes space. Kuuk Thaayorre speakers do not default to their bodies, because their language never taught them to. The details are in the editorial that summarizes this literature, and the original study was published in Psychological Science.

Time: tense, metaphor, and which way the future points

Languages differ in whether they require speakers to mark time grammatically. English insists on it almost constantly. Mandarin has no grammatical tense and relies on context or time words such as "yesterday." Whether that changes how people experience time is an open question.

The Mandarin vertical-time controversy

Boroditsky reported in 2001 that Mandarin speakers were more likely to think about time vertically than English speakers, because Mandarin uses up and down metaphors (shang and xia) more often. The result has had a rough history. Chen's 2007 paper in Cognition, titled "Do Chinese and English speakers think about time differently? Failure of replicating Boroditsky (2001)," found that Chinese speakers actually use horizontal metaphors more often than vertical ones and reported failing to replicate the original result. Two other teams (January and Kako, and Tse and Altarriba) also reported failed replications.

The debate did not end there. Boroditsky, Fuhrman, and McCormick reported in 2011 that Mandarin speakers recognized temporal sequences faster when presented vertically, while English speakers showed no such bias, and Chen and O'Seaghdha later argued the influence of spatiotemporal metaphors on temporal cognition is limited. If you hear that "Mandarin speakers think about time vertically" as settled fact, it is not.

Aymara: the future behind you

In English, the future is ahead and the past is behind. Rafael Núñez and Eve Sweetser reported in 2006 that speakers of Aymara, an Andean language, reverse this. The Aymara word for future is related to the word for "back" or "behind," and the word for past is related to "front." The authors analyzed recorded conversation and gesture from Aymara speakers in northern Chile and found that people gestured ahead when talking about the past and backward when talking about the future.

Their explanation is about knowledge. What you know, you have seen, so it is in front of you. What you cannot see, the future, lies behind. It is a different logic, not a different level of cognitive ability.

Numbers: language as a tool for exactness

Pirahã is an Amazonian language with quantity words that are approximate. According to follow-up work, hói roughly means "about one," hoí means "about two, or a few," and baágiso means "many." There are no exact number words.

Peter Gordon's 2004 study in Science tested Pirahã speakers on matching tasks. Performance on sets larger than three was weak, though errors grew in proportion to set size, which Gordon read as approximate estimation rather than exact counting. Critics, including Daniel Everett, who also worked with the community, objected to some of the materials, such as using batteries as stimuli.

A 2008 follow-up by Michael Frank, Everett, Evelina Fedorenko, and Edward Gibson found that Pirahã speakers could match exact quantities when no memory was involved, but that their matches became approximate when memory was required. The authors argued that number words do not change the underlying representation of quantity. They work as a cognitive technology for keeping track of exact sets across time and space.

That is a good example of what "language shapes thought" can mean in practice. It does not mean the Pirahã cannot perceive that one pile is bigger. It means a technology, number words, makes certain tasks much easier, much as a notebook helps you remember a long list. If you want the full story on how cultures built that technology, how learning actually works covers the role of external tools in cognition.

Framing, blame, and decisions

So far we have looked at how different languages might produce different minds. A second line of research is simpler: wording changes decisions in any language.

The classic framing effect

In 1981, Amos Tversky and Daniel Kahneman asked participants to choose between programs to combat a disease expected to kill 600 people. When the options were framed as lives saved, 72% chose the sure option of saving 200. When identical outcomes were framed as deaths, with 400 certain deaths against a one-in-three chance that nobody dies, only 22% chose the certain option. The details are in the original Science paper, and a longer walkthrough is in the Tversky-Kahneman framing experiment. For how this plays out in everyday risk decisions, see how framing changes risk perception.

This is not a difference between languages. It is a property of human decision-making that works through wording, which is why it belongs in a discussion of language and thought even though it does not show that Russian and English speakers think differently.

Who did it? Agency in English, Spanish, and Japanese

Languages differ in how readily they describe accidents with an agent ("he broke the vase") versus without ("the vase broke"). Caitlin Fausey and Lera Boroditsky found that English speakers used more agentive wording for accidents than Spanish speakers did, and that English speakers also remembered the agents of accidental events better, while the two groups remembered intentional agents about equally well.

A companion study with Japanese speakers found the same pattern. In it, 62 English speakers and 70 Japanese speakers saw filmed events, and in a third experiment, English speakers who first read agentive sentences were better at remembering agents, which hints that language may play a causal role (Fausey et al., 2010). The measured outcome here is memory for who did something, which is narrower than "blame."

The foreign-language effect

If wording biases decisions, what happens when you decide in a second language? Boaz Keysar, Sayuri Hayakawa, and Sun Gyu An reported in 2012 that the framing effect disappeared when participants made choices in a foreign tongue, and they proposed that a foreign language creates distance that favors deliberate thinking over gut reactions.

Albert Costa and colleagues found a similar effect on moral choices. In the footbridge dilemma, where pushing one man would save five, 18% of participants chose to push when using their native language, versus 44% when using a foreign language, in an experiment of 725 people (Spanish speakers using English and English speakers using Spanish). A smaller first experiment with 317 participants found 20% versus 33%. The authors attribute the shift to reduced emotional response in a foreign language (Costa et al., 2014). Our article on why losses feel stronger than gains covers the emotional machinery in the background.

How strong is the evidence? A scorecard

The evidence differs sharply from domain to domain. The table below is a summary of what has been published, with a plain-language read on how settled each claim is. It is a judgment, not a formal meta-analysis.

Domain	Representative finding	Direction of evidence	Main caveat
Color	Faster discrimination across a native category boundary (Russian blues, 2007)	Supportive, with a failed or contradicting follow-up	Effect depends on task, and a later study reported no speed advantage
Color and brain	Stronger effect in the right visual field (2006, 2007)	Replicated in one follow-up	Left-field effects also found, so the asymmetry is partial
Space	Cardinal-direction languages train absolute orientation	Strong, multiple groups	Whether a given language is purely absolute is disputed
Time (east-west)	Kuuk Thaayorre arrange time east to west	Supportive, with a follow-up study	Small communities, so sample sizes are limited
Time (vertical, Mandarin)	Vertical bias in temporal judgments	Mixed	Failed replications reported in 2007 and 2008, later positive results in 2011
Time (Aymara)	Future gestured behind, past ahead	Supportive	Based on conversational and gesture data from a limited sample
Numbers	Exact matching degrades with memory load (Pirahã)	Supportive	Everett and others criticized stimuli and interpretation
Agency and memory	English speakers recall accidental agents better than Spanish or Japanese speakers	Supportive, includes a priming experiment	Measures memory for agents, not broader blame
Decisions	Foreign-language effect on framing and moral choices	Supportive across several groups	Mechanism (emotion vs. processing fluency) is still debated

The pattern across the table is worth noticing. The effects that look most solid are the ones tied to a clear mechanism, such as an organizational habit forced by grammar (compass directions) or a tool that supports memory (number words). The ones that look most fragile are those based on metaphor frequency alone.

What these studies cannot tell you

Even the better studies have limits, and knowing them makes you a better reader of science headlines.

Correlation is not causation. If speakers of language A think differently from speakers of language B, the language might be the cause, or culture, education, environment, and writing system might be. The strongest designs break this link by interfering with language in the moment (as in the verbal interference test) or by priming it directly (as in the agency study). Our explainer on how correlation differs from causation covers why this distinction matters so much.

Small samples and specialized populations. Studies of small communities, such as Pormpuraaw, Aymara speakers in northern Chile, or the Pirahã, are scientifically precious but hard to repeat. Replication is exactly where several well-known results have struggled.

Effect sizes are usually modest. A reaction-time difference of a fraction of a second is real, but it is not the same as "seeing the world differently." Headlines usually blur that line. For a broader look at why published results can disagree, see why scientists disagree so much.

Task matters. Several effects appear only when people must hold information in language-based memory or are under time pressure. Remove the pressure and the differences often shrink.

What this means for translation, learning, and design

Translation

If language biases attention and memory, no translation is perfectly neutral. That does not make translation impossible, because people can express almost any idea in almost any language with enough words. It means a translator is making choices about what to foreground, such as agency, sequence, or direction. Machine translation faces the same issue, which is part of why AI translation tools still need human review for sensitive text.

Learning another language

Learning a second language adds vocabulary and may shift habits of attention. Whether it also gives general cognitive benefits is less clear. A 2018 meta-analysis in Psychological Bulletin by Minna Lehtonen and colleagues pooled 891 effect sizes from 152 studies of adults and found that a small apparent bilingual advantage in some executive-function domains disappeared after correcting for publication bias. The authors concluded the evidence does not systematically support a general bilingual benefit in cognitive control. If you are weighing options, how to learn a new language with AI discusses what works in practice, and the neuroscience of learning Elvish looks at what invented languages show about the brain.

Bilingual minds

Studies of bilinguals suggest that both languages tend to be active at the same time when one is being used, a pattern Judith Kroll and colleagues summarized with the phrase "language selectivity is the exception, not the rule." Selecting one language therefore involves active control. Whether that amounts to "thinking in two languages at once" is a definition question more than an empirical one.

Communication and design

Wording matters. Health messages, product labels, and survey questions can all steer answers through framing, which is why careful researchers test several phrasings. A related tendency is covered in cognitive biases explained, and the broader hypothetical is explored in what if language completely shaped thought.

Common misconceptions

"Eskimo languages have 100 words for snow." This popular claim is usually told as proof of linguistic relativity, and it is not one. Nothing in the studies above depends on it, and the number is widely regarded as exaggerated, so treat it as folklore, not evidence.

"If you have no word for it, you cannot think it." That is the strong form, which researchers reject. Russian and English speakers can both see two shades of blue as different. What differs is how quickly and easily.

"Language is the same thing as thought." Much thinking is nonlinguistic: spatial reasoning, emotion, music, pattern recognition. Language is one powerful tool among several.

"Some languages are better for thinking." There is no evidence that any natural language is superior overall. Languages differ in what they make easy, such as precise number words for arithmetic, not in total expressive power.

"A single experiment proves it." Several of the best-known findings in this field have both supporting and contradicting studies. Treat any single result, including the ones above, as one data point.

The takeaway

Language is best understood as a cognitive tool that makes some mental operations cheaper and others costlier. Words for exact numbers make exact counting across time possible. A grammar that forces you to say north and south makes you keep track of north and south. A habit of agentive phrasing may make you remember who did what. None of that traps you, but all of it nudges you.

The most useful lesson is about evidence, not about any one language. The strongest results are the ones where a clear mechanism links a grammatical habit to a measurable task, and the weakest are the ones that rest on a metaphor alone. When you read the next claim that some language makes its speakers see the world differently, ask what the mechanism is, what the effect size was, and whether anyone has tried to repeat it.

Frequently asked questions

Q: Can you think without language? A: Yes. Spatial reasoning, emotional experience, music, and pattern recognition do not require words. The experiments above show language influences some forms of thought, not that all thought depends on it. Even in the Russian blues study, the effect disappeared only when verbal interference occupied the language system, which implies speakers can see the colors regardless.

Q: Is one language better than another for thinking? A: No evidence supports that. Languages differ in what they make easy, such as compass directions in Guugu Yimithirr or exact number words in English, but no natural language has been shown to give its speakers a general advantage in reasoning.

Q: Does thinking in a second language change your decisions? A: It can. Keysar and colleagues reported in 2012 that the framing effect disappeared in a foreign language, and Costa and colleagues found that 44% of participants chose the utilitarian push in a foreign language versus 18% in their native one in a 725-person experiment. The proposed explanation is reduced emotional reactivity, though the mechanism is still debated.

Q: Do bilinguals have better cognitive abilities? A: The evidence is weak. A 2018 meta-analysis of 152 adult studies found that a small apparent advantage in some executive-function domains vanished after correcting for publication bias.

Q: Can you think in multiple languages at once? A: Both of a bilingual's languages tend to be active whenever one is used, so the other must be suppressed or controlled. Bilinguals also code-switch within sentences. Whether that counts as thinking in two languages at once depends on how you define thinking.

Q: Was Whorf right about the Hopi? A: His claim that Hopi has no way to talk about time was challenged by Ekkehart Malotki in 1983, who reported finding no support for it in 1980s Hopi speech or historical documents. Some scholars, such as John Lucy, argue Malotki misread Whorf, so the specifics remain debated, but the claim is no longer treated as established.

Sources

• Linguistic relativity (overview of history, Brown and Lenneberg, Malotki, replication notes)
• Winawer et al. (2007), Russian blues reveal effects of language on color discrimination, PNAS
• Drivonikou et al. (2007), Whorfian effects in the right visual field, PNAS
• Majid et al. (2004), Can language restructure cognition? The case for space
• Fausey et al. (2010), Constructing Agency: The Role of Language
• Costa et al. (2014), Your Morals Depend on Language, PLOS ONE
• Lehtonen et al. (2018), Bilingualism and executive functioning, Psychological Bulletin
• Tversky and Kahneman (1981), The Framing of Decisions, Science

---

Related Reading

• How Bayesian Thinking Can Make You Dramatically Better at Reasoning
• How Many Humans Have Ever Lived? The Statistics of Human Existence
• How Does a Refrigerator Work? The Cool Science Behind Everyday Chill
• How the Large Hadron Collider Works, and What It's Searching For
• Why You Can't Feel the Earth Spinning at 1,600 km/h